Question about 3D obejct detection results

mileyan / Pseudo_Lidar_V2

(ICLR) Pseudo-LiDAR++: Accurate Depth for 3D Object Detection in Autonomous Driving

MIT License

594 stars 124 forks source link

Question about 3D obejct detection results #4

Open gujiaqivadin opened 4 years ago

gujiaqivadin commented 4 years ago

Hello,mileyan! Thanks for sharing your great code about Pseudo LiDAR and its ++ version. I am testing your output pseudo lidar on second/pointpillars and other voxel-based 3d object detection methods but found the results of them are relatively poor compared to frustumpointnet/PointRCNN and other methods you mentioned in your paper 1 and 2. I wonder have you tested your lidar on these voxel-based 3d obejct detection methods? And how will it turn out to be finally? Thanks a lot!

mileyan commented 4 years ago

Hi, thanks for your interest. AVOD and PIXOR are voxel-based methods. We have presented the results of them in our paper. Thanks for pointing it out. I haven't got a chance to test PL on second/pointpillars. I will try PL on these models. If you have any interesting find, please let me know.